Early prediction of Mycobacterium tuberculosis transmission clusters using supervised learning models.

Area of Science:

• Public health microbiology
• Infectious disease epidemiology
• Machine learning in healthcare

Background:

• Identifying individuals with tuberculosis (TB) who are likely to transmit the disease is crucial for effective public health strategies and disease prevention.
• Current methods may not sufficiently identify high-transmission risk individuals within Mycobacterium tuberculosis transmission clusters.

Purpose of the Study:

• To develop and evaluate supervised machine learning models for predicting individuals with TB who are likely to be the source of new transmission events within clusters.
• To identify key demographic and clinical factors associated with TB transmission potential.

Main Methods:

• Trained classification models, specifically balanced random forests, using demographic and disease data from individuals with TB.
• Utilized data from the first sampled isolates in Mycobacterium tuberculosis transmission clusters to predict cluster growth.
• Evaluated model performance using Area Under the Curve (AUC) metrics.

Main Results:

• Supervised learning models, particularly balanced random forests, demonstrated good performance (AUCs ≥ 0.75) in predicting individuals associated with TB cluster growth.
• Key predictors identified included host demographics, site of infection, TB lineage, and age at diagnosis.
• The models can effectively identify individuals with a higher likelihood of contributing to TB transmission.

Conclusions:

• A machine learning framework can be developed to predict potential TB cluster growth.
• This predictive capability allows for early assessment and prioritization of individuals for enhanced follow-up interventions.
• The ultimate aim is to reduce TB transmission chains and improve public health outcomes.